Skip to main content
Monk is a DevOps agent that plugs into your coding agent. You say what you want — /monk deploy this project — and Monk handles the rest: containers, infrastructure, databases, networking, and monitoring, on your own cloud. This page takes you from install to a running app.

1. Install Monk in Your Agent

Monk installs as a plugin. One install gets you everything: a local monk-agent companion, the MCP connection to your agent, and the monk skill. There is no separate app to run and nothing to configure by hand.
The plugin installs monk-agent under ~/.monk/bin and connects your agent to its MCP endpoint on 127.0.0.1:7419. On first use it also installs the Monk CLI and daemon (monk, monkd) through your system package manager — Homebrew on macOS, apt or dnf on Linux, a dedicated Ubuntu WSL distro on Windows. You need network access to GitHub and https://get.monk.io/stable.
Run these commands inside Claude Code:
Reload the current session:
Then run /mcp — Monk should appear as an MCP server. If Claude Code asks you to authenticate, complete the browser sign-in flow from /mcp.
Your coding agent should call Monk’s MCP tools for infrastructure and runtime operations — not run direct monk ... shell commands against live Monk-managed infrastructure. Any other MCP-capable agent can connect to http://127.0.0.1:7419/mcp, but the hosts above are the supported, tested paths.

2. Sign In

Check that everything is wired up:
Monk inspects the local runtime, installs anything still missing, and walks you through the browser sign-in. You need a Monk account: the easiest way is to sign up at monk.io/dashboard, or let the plugin prompt you — either way it’s a 5-day free trial, no credit card required.

3. First Deployment

Open your project, then tell your coding agent:
Monk analyzes the code, builds containers, provisions cloud infrastructure, and wires everything together. You will be asked to pick a team, enter credentials in the local Monk dashboard (never in chat), and approve the infrastructure. Nothing else. A typical first deployment finishes in five to fifteen minutes. When it is done you get live URLs, service health, and an estimated monthly cost.

First deployment walkthrough

Step-by-step guide with example prompts and common scenarios

4. Autonomous Operations

After your first deployment, Monk keeps your system running and gives you tools to manage it without writing infrastructure code. Watcher monitors your services around the clock and alerts you on Slack when something needs attention. CI/CD can be set up with a single prompt so every push triggers a new deployment. And you can scale, migrate, or reconfigure services at any time through natural language.

How autonomous operations work

What Monk handles on its own vs. what it asks you

Watcher setup

Continuous monitoring and Slack alerts

CI/CD setup

Auto-deploy on every code push

Prompting guide

Tips for getting the most out of Monk

5. Connect Your Cloud

Monk follows a bring-your-own-infrastructure model. Everything runs on your cloud accounts. Monk asks for credentials when it needs them, but you can also set them up ahead of time.

AWS

IAM access key and secret key

Google Cloud

Service account JSON key

Microsoft Azure

Service principal with client secret

DigitalOcean

Personal access token

Hetzner

API token
Monk also integrates with service providers like MongoDB Atlas, Redis Cloud, Netlify, Vercel, and Auth0. Credentials for these are requested automatically when you first use them.

All credentials and permissions

Full reference for every provider, including minimum IAM policies

6. Getting Help

If something goes wrong, start by asking Monk directly — it has full context about your project and infrastructure. For bugs, say /monk report a bug right in the chat — Monk collects environment details automatically and sends them straight to the engineering team (it can also file reports on its own when it hits a problem). You can also use the bug icon in the top-right corner of the Monk dashboard.

Help & support channels

Bug reports, feature requests, community forum, and direct contact

Troubleshooting

Common setup and connection issues

Updating

Uninstall

First remove the plugin from the host: Host uninstall removes the plugin, but not the local companion. To remove it, run the cleanup script from the plugin repository:
On Windows:
To also remove the Monk runtime components (CLI and daemon), add the runtime option: --runtime on macOS/Linux, -Runtime on Windows.